调整ggplot中plot.background边距，解决图表背景过大问题

问题根源 #

1. 误用scale_y_discrete()：prob是连续数值变量，却用了离散刻度设置，导致y轴方向生成多余空白
2. coord_flip()配合单个离散x类别时，ggplot默认保留了大量冗余空间
3. 用标题margin临时调整的方式治标不治本，未从图表布局本身解决边距问题

具体解决方案 #

• 替换scale_y_discrete()为scale_y_continuous()，设置expand = c(0, 0)彻底消除y轴上下空白
• 调整x轴expand参数，进一步压缩左右方向的冗余空间
• 使用theme(plot.margin)精准控制整个图表的上下左右边距，替代标题的临时margin设置
• 若用cowplot拼接图表，通过plot_grid()的rel_heights和padding参数控制图间距

修改后的完整代码 #

library(dplyr)
library(ggplot2)

# Data
win_prob <- structure(list(winner = c("draw", "team_a", "team_b"), prob = c(0.1374, 
0.8412, 0.0214), team_name = c("Draw", "BYU", "Stanford")), row.names = c(NA, 
-3L), class = "data.frame")

# Plot
p1 <- win_prob |>
  mutate(dummy = "dummy") |>
  ggplot(aes(x = dummy, y = prob, fill = factor(winner, levels=c("team_b", "draw", "team_a")))) +
  geom_col(width = 0.02, alpha = 0.50) +
  geom_text(aes(label = paste0(round(prob*100, 0), "%")), position = position_stack(vjust = 0.5)) +
  scale_fill_manual(values = c("#8C1515", "grey", "#002E5D")) +
  coord_flip() +
  # 修正x轴expand，压缩左右空白
  scale_x_discrete(expand = c(0.01, 0)) +
  # 替换为连续y轴，消除上下空白
  scale_y_continuous(expand = c(0, 0)) +
  theme(
    legend.position = "none",
    # 精准控制图表整体边距，单位为mm
    plot.margin = margin(t = 10, r = 20, b = 10, l = 10, unit = "mm"),
    plot.title = element_text(size = 40),
    plot.subtitle = element_text(size = 28)
  ) +
  annotate("text", x = 1.02, y = 0.0, hjust = 0, label = win_prob$team_name[2], size = 8) +
  annotate("text", x = 1.02, y = 1.0, hjust = 1, label = win_prob$team_name[3], size = 8) +
  annotate("text", x = 1.02, y = 0.5, hjust = 0.5, label = "Draw", size = 8) +
  labs(title = "BYU vs Stanford Win Probability",
       subtitle = "based on xG simulation")

# 若用cowplot拼接示例（假设p2是第二个图表）
# library(cowplot)
# plot_grid(p1, p2, ncol = 1, rel_heights = c(1, 1), padding = 0.05)

补充说明 #

• plot.margin的四个参数分别对应上、右、下、左的边距，可根据需求调整数值
• 若需要保留theme_void()，只需在theme()里补充plot.margin即可，无需保留其他主题元素
• cowplot拼接时，padding参数控制图之间的相对间距，rel_heights可调整各图的高度比例

内容的提问来源于stack exchange，提问作者seansteele